diff --git a/main.tex b/main.tex index a636888ede7c74e40623bbbf465d65e1faa86b97..9dbddad363240121706701aad2442dd51093882b 100644 --- a/main.tex +++ b/main.tex @@ -331,7 +331,7 @@ We solve this problem by decoupling the gradient averaging from the weight avera % To regenerate the figure, from directory results/scaling -% python ../../../learn-topology/tools/plot_convergence.py 10/mnist/clique-ring/all/ ../mnist/clique-ring/all/ 1000/mnist/clique-ring/all/ --labels '10 nodes bsz=1280' '100 nodes bsz=128' '1000 nodes bsz=13' --legend 'lower right' --yaxis test-accuracy --save-figure ../../figures/d-cliques-mnist-scaling-clique-ring-cst-updates.png --ymin 80 --add-min-max +% python ../../../learn-topology/tools/plot_convergence.py 10/mnist/clique-ring/all/2021-03-13-18:22:01-CET ../mnist/clique-ring/all/2021-03-10-18:14:35-CET 1000/mnist/clique-ring/all/2021-03-13-18:22:36-CET --labels '10 nodes bsz=1280' '100 nodes bsz=128' '1000 nodes bsz=13' --legend 'lower right' --yaxis test-accuracy --save-figure ../../figures/d-cliques-mnist-scaling-clique-ring-cst-updates.png --ymin 80 --add-min-max \begin{subfigure}[b]{0.7\textwidth} \centering \includegraphics[width=\textwidth]{figures/d-cliques-mnist-scaling-clique-ring-cst-updates} diff --git a/results/cifar10/random-10-diverse-unbiased-gradient-uniform-init/experiments.sh b/results/cifar10/random-10-diverse-unbiased-gradient-uniform-init/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..5f0ba427898615df2ad16c6ccf8e47b3734fe180 --- /dev/null +++ b/results/cifar10/random-10-diverse-unbiased-gradient-uniform-init/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 20 + ' +LRS=' + 0.002 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 1 1 1 1 1 1 1 1 1 1 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0.9 --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --single-process --nb-logging-processes 10 --dataset cifar10 --model gn-lenet --accuracy-logging-interval 10 --validation-set-ratio 0.5 --unbiased-gradient --initial-averaging + done; +done; + diff --git a/results/cifar10/random-10-diverse-unbiased-gradient/experiments.sh b/results/cifar10/random-10-diverse-unbiased-gradient/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..c8480cbef17721621d526c47d291b5b0b8a39601 --- /dev/null +++ b/results/cifar10/random-10-diverse-unbiased-gradient/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 20 + ' +LRS=' + 0.002 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 1 1 1 1 1 1 1 1 1 1 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0.9 --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --single-process --nb-logging-processes 10 --dataset cifar10 --model gn-lenet --accuracy-logging-interval 10 --validation-set-ratio 0.5 --unbiased-gradient + done; +done; + diff --git a/results/cifar10/random-10-diverse/experiments.sh b/results/cifar10/random-10-diverse/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..deff5d63c3a3c896b991fb4b4da41c1828dedcb5 --- /dev/null +++ b/results/cifar10/random-10-diverse/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 20 + ' +LRS=' + 0.002 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 1 1 1 1 1 1 1 1 1 1 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0.9 --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --single-process --nb-logging-processes 10 --dataset cifar10 --model gn-lenet --accuracy-logging-interval 10 --validation-set-ratio 0.5 + done; +done; + diff --git a/results/cifar10/random-10/experiments.sh b/results/cifar10/random-10/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..dc2eeb99061ff42500ffbf66909c4c701e229c51 --- /dev/null +++ b/results/cifar10/random-10/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 20 + ' +LRS=' + 0.002 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 1 1 1 1 1 1 1 1 1 1 --dist-optimization d-psgd --topology random-10 --metric dissimilarity --learning-momentum 0.9 --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --single-process --nb-logging-processes 10 --dataset cifar10 --model gn-lenet --accuracy-logging-interval 10 --validation-set-ratio 0.5 + done; +done; + diff --git a/results/mnist/random-10-diverse-unbiased-grad-uniform-init/experiments.sh b/results/mnist/random-10-diverse-unbiased-grad-uniform-init/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..238be57998387663b2f15f5959b5c505c633085f --- /dev/null +++ b/results/mnist/random-10-diverse-unbiased-grad-uniform-init/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 128 + ' +LRS=' + 0.1 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --unbiased-gradient --initial-averaging + done; +done; + diff --git a/results/mnist/random-10-diverse-unbiased-grad/experiments.sh b/results/mnist/random-10-diverse-unbiased-grad/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..b461259aa3698395ba1e41710dc859821d47fa8f --- /dev/null +++ b/results/mnist/random-10-diverse-unbiased-grad/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 128 + ' +LRS=' + 0.1 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --unbiased-gradient + done; +done; + diff --git a/results/mnist/random-10-diverse/experiments.sh b/results/mnist/random-10-diverse/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..80e6b7154b383e898dde07f10526b0ad12b28ff5 --- /dev/null +++ b/results/mnist/random-10-diverse/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 128 + ' +LRS=' + 0.1 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology greedy-diverse-10 --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear + done; +done; + diff --git a/results/mnist/random-10/experiments.sh b/results/mnist/random-10/experiments.sh new file mode 100755 index 0000000000000000000000000000000000000000..ba5dc2ab6e3f8d6a0a258f95b5134224313e785d --- /dev/null +++ b/results/mnist/random-10/experiments.sh @@ -0,0 +1,14 @@ +#!/usr/bin/env bash +TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS +BSZS=' + 128 + ' +LRS=' + 0.1 + ' +for BSZ in $BSZS; + do for LR in $LRS; + do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology random-10 --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear + done; +done; +